System tools

Collecting the contents of every .htaccess file of a website

Since we use NGINX and not Apache, the contents of the .htaccess files have to be checked to know what needs to be implemented as an equivalent NGINX rule:

bash

collect-htaccess.sh <user> [subdirectory]

Measuring the memory footprint of a gunicorn worker

Workers are forked from the master and share part of their pages: RSS therefore counts the same memory several times. PSS is the figure to read.

bash

for pid in $(cat /sys/fs/cgroup/system.slice/<service>.service/cgroup.procs); do
    printf '%6s %5s MB  %s\n' "$pid" \
        "$(( $(awk '/^Pss:/{print $2}' /proc/$pid/smaps_rollup) / 1024 ))" \
        "$(tr '\0' ' ' < /proc/$pid/cmdline | cut -c1-60)"
done

The first PID is the master, the following ones are the workers. The total for the service, as a cross-check:

bash

systemctl show -p MemoryCurrent <service>.service

Measure after real traffic, not right after startup: the footprint grows with loaded modules and caches. The value of one worker, rounded up, goes into per_worker_mb of the site's hiera declaration, which is used to compute the number of workers. Repeat after a major version upgrade or after adding a heavy dependency.

See also

• Upgrading Debian 12 to Debian 13 (trixie).
• Converting a network configuration to netplan.
• Virtualmin on the command line.